Review: Buffalo 2×1GB PC2-8000 Firestix
12.19.07 - 07:18pm
It seems to be very late in the DDR2 life cycle to be reviewing a set of high-end sticks considering however DDR2 will live on for at least another year in the enthusiast market so fear not, this review was not done in vain. DDR2 on Intel motherboards is still capable of performing within a very close margin of current DDR3 motherboards and AMD will be solely using DDR2 until late 2008/early 2009. Currently unless you have very deep pockets or an unlimited budget then high-end DDR3 is out of your price-range which leaves you with what some would consider to be old technology. I think however that when you see the results of the pinnacle of the DDR2 era you may be swayed if you were part of the previously mentioned concept. Without further ado, here’s the review of the Buffalo Technology PC2-8000 Firestix, clad in flashy red heatsinks.
To be honest, prior to purchasing these sticks I hadn’t heard of Buffalo Technology but rest assured they are now one of the few companies that I hold in high regards. Buffalo Technology currently manufacturers a wide range of products within the memory, storage, and networking segments of the industry. Browsing over some of there non-memory products reveals a very diverse product line with multiple objects standing out, most noticeably their line of NAS boxes. That TeraStation looks positively divine but even their small MiniStation portable USB drive features a crisp and clean design. Besides these products, Buffalo also manufactures four separate lines of desktop and mobile DIMMs ranging from OEM products like their Select and TechWorks lines, a budget Certified line, and then their premier Firestix line. Currently the Firestix line contains both DDR and DDR2 DIMMs with speeds ranging from 400MHz to 433MHz with DDR and 800MHz to 1200MHz with DDR2. Today I will be reviewing the 2 gigabyte 1000MHz kit which comes in a pair of matched 1GB DIMMs.
I like to believe that the designers of memory heatsinks all hold large conferences where they all decide who has the best looking heatsinks. At these conferences lots of awards are handed out, everyone is cheerful, and there are hundreds of pedestals with various prototypes that may one day get slapped onto a SMT covered PCB and sold to the consumer. Given the little amount of room you have to work with in designing DIMM heatsinks, it’s no surprise that most designs are the standard stamped and anodized aluminum plate with steel clips. The Buffalo Firestix feature a very vivid fiery red anodized heatsink with the word Firestix blazoned across the the aluminum and a unique bit of abstract art behind the name. As you can see below, the sticks are attractive and if I don’t end up selling them I think I’ll frame them next to my 10th Anniversary Crucial sticks which sit next to the Firestix. While looks are only a small portion of the product, when designing a system for a show it can make or break a purchase.
Within the heatspreaders sit 16 Micron D9GKX modules sitting on a 6 layer PCB. Micron D9GKX modules are the current champ when it comes to hitting stratospheric clocks but care must be taken when overclocking Micron based sticks. Since the introduction of the D9GMH and D9GKX modules users have noted that modules tend to degrade and die when voltages of 2.2v or higher are used. I personally run my sticks 24/7 at 1.95v which is enough voltage to provide the overhead for decent clocks and timings. That being said, D9GKX modules have been pushed with 2.6v+ with clockspeeds in excess of 1400MHz so these Firestix have much promise. Besides being capable of clocking well with lots of voltage, D9GKX is also known for hitting decent clocks with minimal voltages, as will be seen in a bit.
For this review I”ll be using my freshly updated testbed. For this particular review I was stuck with using air-cooling as my cascade is currently in pieces and I didn’t have a sufficient single-stage to handle a Q6600, so all tests were done around 2400MHz. The P35 chipset is known to clock DDR2 relatively well and I was rather impressed with how it clocked these sticks but a 780i or X38 based motherboard would have had the ability to possibly push this sticks further so keep that in mind. As always remember that why results are only a single sample from the 1000MHz Firestix so it’s possible to get completely different results on your setup but for the most part this issue is a thing of the past except for the rare exceptional handpicked sticks.
Hardware
- Processor: Intel Q6600, G0 stepping, clocks up to 490FSB and 4050MHz on air.
- Motherboard: MSI P35 Neo2-FR, clocks DDR2 past 1200MHz
- Videocard: eVGA 7900 GTO @ stock clocks
- Harddrive:Seagate 300GB 7200.10
- Powersupply: OCZ 700w GameXStream
- Cooling: Scythe Ninja Rev B with 120mm 110cfm fan, 120mm 110cfm fan over northbridge and ram
Voltages
- CPU: 1.20v
- vDIMM: 1.80v, 2.10v, 2.30v
- VTT: 1.425v
- NorthBridge: 1.45v
- SouthBridge I/O: 1.5v
- SouthBridge: 1.05v
I decided to do three stages of testing with three different sub-stages within each primary stage. The primary stages were tests with the memory voltages set to 1.8v, 2.1v, and 2.3v. The sub-stages were tests with the memory timings set to 3-3-3-9, 4-4-4-12, and 5-5-5-15. I did this so that you could easily see how capable these sticks are. To test for stability at a specific setting I would run four instances of SuperPi calculating to 32 million digits, I find that typically 32M will fail instantly if there is a slight amount of memory instability so this gives a pretty accurate quick-test. I imagine for 24/7 settings for each given result I would drop the clocks 5% or so to insure stability with the given volts and timings. For each specific maximum value I then recorded the integer and floating point bandwidth as reported by Sandra 2008 while the bandwidth efficiency was also recorded. For the most part the tests advanced in a predictable manner but there were a few flukes, as will be seen in the graphs below.
As you can see the sticks were very capable when operating with 1.8v and 2.1v but for some reason the 2.3v performance was a let-down. I am positive though that the culprit was the memory divider that was used for the final 2.3v 5-5-5-15 test. All the other tests utilized 1:1, 2:3, or 5:6 dividers but the last test used the 1:2 divider which apparently caused something to cut bandwidth by 2GBps. That final test did have the lowest FSB due to the 1:2 divider, I believe with the 2:3 divider it would have easily resulted in closer to 8000 GBps of bandwidth in both the integer and floating point tests. Most impressive of all I would have to say was the 1032MHz operating with 1.8v and 5-5-5-15 timings. With a better clocking board I could easily see 1250-1300MHz capable from these sticks with 2.3v and with even higher voltage perhaps 1125MHz 4-4-4-12 performance would be possible.
The Buffalo Firestix offer what I would consider to be stellar performance from a upper-midrange memory kit. Currently I cannot seem to find any retailers besides Buffalo Technology who are charging $277 for this kit so it’s hard to suggest buying directly from them when there are some low-end DDR3 kits available in that price-range. I originally purchased these sticks roughly a year ago and just recently received this particular set two weeks ago fresh from RMA so I’m not sure if they have been partially discontinued but I sure hope not. Buffalo Technology has a fantastic set of sticks in this kit and from my experience their customer service was top notch and I would definitely purchase from Buffalo again. All being said, if you can find this same kit in the $90 to $120 price-range I would suggest jumping on them as they would be a good buy.
Final Score: 9/10
Pros
- Extreme Low and Medium Voltage Performance
- Sleek Memory Heatsinks
- Excellent Customer Service
Cons
- Hard to Find
- Expensive when Purchased from Buffalo Technology
Speak Your Mind